Visualization All Studies Data

knitr::opts_chunk$set(echo = TRUE)
knitr::opts_chunk$set(message = TRUE)
knitr::opts_chunk$set(warning = FALSE)

library(haven)
library(tidyverse)

## ── Attaching packages ─────────────────────────────────────── tidyverse 1.3.2 ──
## ✔ ggplot2 3.3.6      ✔ purrr   0.3.5 
## ✔ tibble  3.1.8      ✔ dplyr   1.0.10
## ✔ tidyr   1.2.1      ✔ stringr 1.4.1 
## ✔ readr   2.1.3      ✔ forcats 0.5.2 
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag()    masks stats::lag()

library(dplyr)
library(ggplot2)
library(plotly)

## 
## Attaching package: 'plotly'
## 
## The following object is masked from 'package:ggplot2':
## 
##     last_plot
## 
## The following object is masked from 'package:stats':
## 
##     filter
## 
## The following object is masked from 'package:graphics':
## 
##     layout

Merging Data Across Studies

#Data Import
dat2 <- read_sav("~/Desktop/Academics/PhD Research/U.S. RDE/2nd Run/AllData.sav")

dat2 <- dat2 %>%
  rename(pro1 = "Q5",
         pro2 = "Q6",
         pro3 = "Q8",
         pre1 = "Q11",
         pre2 = "Q12",
         pre3 = "Q13",
         DV = "DVTotalAct") %>%
  select(pro1, pro2, pro3, pre1, pre2, pre3,
         Race_1, Race_2, Race_3, Race_4, Race_5, Race_6, Race_7, Race_8, Race_9,
         Gender, Politics, SES, Condition, DV) %>%
  mutate(Survey = 2)

#Data Import
dat4 <- read_sav("~/Desktop/Academics/PhD Research/U.S. RDE/Resonance:OMT Version/1st Study/AllDataOMT.sav")

dat4 <- dat4 %>%
  rename(pro1 = "Q5",
         pro2 = "Q6",
         pro3 = "Q8",
         pre1 = "Q11",
         pre2 = "Q12",
         pre3 = "Q13",
         DV = "DVTotal") %>%
  select(pro1, pro2, pro3, pre1, pre2, pre3,
         Race_1, Race_2, Race_3, Race_4, Race_5, Race_6, Race_7, Race_8, Race_9,
         Gender, Politics, SES, Condition, DV) %>%
  mutate(Survey = 4)

dat5 <- read_sav("~/Desktop/Academics/PhD Research/U.S. RDE/Resonance:OMT Version/2nd Study/AllData.sav")

dat5 <- dat5 %>%
  rename(pro1 = "Q5",
         pro2 = "Q6",
         pro3 = "Q8",
         pre1 = "Q11",
         pre2 = "Q12",
         pre3 = "Q13",
         DV = "DVTotal") %>%
  select(pro1, pro2, pro3, pre1, pre2, pre3,
         Race_1, Race_2, Race_3, Race_4, Race_5, Race_6, Race_7, Race_8, Race_9,
         Gender, Politics, SES, Condition, DV) %>%
  mutate(Survey = 5)

alldat <- rbind(dat2, dat4, dat5)

Tidying Data

#Race_9 = Prefer not to say
#Coding racial variables

alldat <- alldat %>%
  mutate(Race_1 = ifelse(is.na(Race_1), 0, Race_1),
         Race_2 = ifelse(is.na(Race_2), 0, Race_2),
         Race_3 = ifelse(is.na(Race_3), 0, Race_3),
         Race_4 = ifelse(is.na(Race_4), 0, Race_4),
         Race_5 = ifelse(is.na(Race_5), 0, Race_5),
         Race_6 = ifelse(is.na(Race_6), 0, Race_6),
         Race_7 = ifelse(is.na(Race_7), 0, Race_7),
         Race_8 = ifelse(is.na(Race_8), 0, Race_8),
         racesort = (Race_1 + Race_2 + Race_3 + Race_4 + Race_5 + Race_6 + Race_7),
         na = ifelse(racesort < 1, 4, 0),
         White = ifelse(Race_1 == 0 | racesort == 0 | racesort > 1, 0, 1),
         Black = ifelse(Race_2 == 1, 2, 0),
         RaceGroup = (White + Black + na),
         RaceGroup = ifelse(RaceGroup < 1, 3, RaceGroup),
         RaceGroup = ifelse(RaceGroup > 3, NA, RaceGroup))
# finally the race sorting works accurately! Urghh too forever; I wonder if there's a better way       

        
# Make factor variables
alldat$Gender <- factor(alldat$Gender,
                        levels = c(1,2,3,4),
                        labels = c("Male", "Female", "Non-binary", "Transgender"))
alldat$Politics <- factor(alldat$Politics,
                          levels = c(1, 2, 3),
                          labels = c("Democrat", "Independent", "Republican"))
alldat$SES <- factor(alldat$SES,
                     levels = c(1,2,3,4,5),
                     labels = c("$10k and lower", "$10-41k", "$41-87k", "$87-165k", "$165k and above"))
alldat$Condition <- factor(alldat$Condition,
                        levels = c(1,2),
                        labels = c("Promotion Frame", "Prevention Frame"))
alldat$Survey <- as.factor(alldat$Survey)
alldat$RaceGroup <- factor(alldat$RaceGroup,
                           levels = c(1,2,3),
                           labels = c("White", "Black", "non-Black POC"))

#remove the individual racial columns
#also calculate average mindtypes for descriptives
cleandat <- alldat %>%
  select(!starts_with("Race_") & !racesort & !na & !White & !Black) %>%
  mutate(promotion = (pro1 + pro2 + pro3)/3,
         prevention = (pre1 + pre2 + pre3)/3,
         difference = promotion - prevention)

Data Visualization

Descriptives Statistics

summary(cleandat)

##       pro1            pro2            pro3           pre1            pre2      
##  Min.   :1.000   Min.   :1.000   Min.   :1.00   Min.   :1.000   Min.   :1.000  
##  1st Qu.:3.000   1st Qu.:3.000   1st Qu.:3.00   1st Qu.:2.000   1st Qu.:3.000  
##  Median :3.000   Median :3.000   Median :4.00   Median :3.000   Median :4.000  
##  Mean   :3.137   Mean   :3.093   Mean   :3.29   Mean   :2.809   Mean   :3.509  
##  3rd Qu.:4.000   3rd Qu.:4.000   3rd Qu.:4.00   3rd Qu.:4.000   3rd Qu.:4.000  
##  Max.   :4.000   Max.   :4.000   Max.   :4.00   Max.   :4.000   Max.   :4.000  
##       pre3               Gender           Politics                SES     
##  Min.   :1.000   Male       :452   Democrat   :396   $10k and lower :179  
##  1st Qu.:3.000   Female     :544   Independent:365   $10-41k        :328  
##  Median :4.000   Non-binary : 21   Republican :234   $41-87k        :324  
##  Mean   :3.441   Transgender:  6   NA's       : 34   $87-165k       :137  
##  3rd Qu.:4.000   NA's       :  6                     $165k and above: 23  
##  Max.   :4.000                                       NA's           : 38  
##             Condition         DV        Survey          RaceGroup  
##  Promotion Frame :517   Min.   :1.000   2:300   White        :492  
##  Prevention Frame:511   1st Qu.:3.286   4:300   Black        :265  
##  NA's            :  1   Median :3.792   5:429   non-Black POC:246  
##                         Mean   :3.679           NA's         : 26  
##                         3rd Qu.:4.286                              
##                         Max.   :5.000                              
##    promotion       prevention      difference      
##  Min.   :1.000   Min.   :1.000   Min.   :-3.00000  
##  1st Qu.:2.667   1st Qu.:3.000   1st Qu.:-0.33333  
##  Median :3.333   Median :3.333   Median : 0.00000  
##  Mean   :3.173   Mean   :3.253   Mean   :-0.07969  
##  3rd Qu.:3.667   3rd Qu.:3.667   3rd Qu.: 0.33333  
##  Max.   :4.000   Max.   :4.000   Max.   : 2.66667

sd(cleandat$pro1)

## [1] 0.931339

sd(cleandat$pro2)

## [1] 0.9183832

sd(cleandat$pro3)

## [1] 0.8940458

sd(cleandat$pre1)

## [1] 1.030321

sd(cleandat$pre2)

## [1] 0.8638674

sd(cleandat$pre3)

## [1] 0.8708939

sd(cleandat$promotion)

## [1] 0.7817941

sd(cleandat$prevention)

## [1] 0.7175935

sd(cleandat$DV)

## [1] 0.8323799

Heatmap

library(ComplexHeatmap)

## Loading required package: grid

## ========================================
## ComplexHeatmap version 2.13.3
## Bioconductor page: http://bioconductor.org/packages/ComplexHeatmap/
## Github page: https://github.com/jokergoo/ComplexHeatmap
## Documentation: http://jokergoo.github.io/ComplexHeatmap-reference
## 
## If you use it in published research, please cite either one:
## - Gu, Z. Complex Heatmap Visualization. iMeta 2022.
## - Gu, Z. Complex heatmaps reveal patterns and correlations in multidimensional 
##     genomic data. Bioinformatics 2016.
## 
## 
## The new InteractiveComplexHeatmap package can directly export static 
## complex heatmaps into an interactive Shiny app with zero effort. Have a try!
## 
## This message can be suppressed by:
##   suppressPackageStartupMessages(library(ComplexHeatmap))
## ========================================

## 
## Attaching package: 'ComplexHeatmap'

## The following object is masked from 'package:plotly':
## 
##     add_heatmap

library(circlize)

## ========================================
## circlize version 0.4.15
## CRAN page: https://cran.r-project.org/package=circlize
## Github page: https://github.com/jokergoo/circlize
## Documentation: https://jokergoo.github.io/circlize_book/book/
## 
## If you use it in published research, please cite:
## Gu, Z. circlize implements and enhances circular visualization
##   in R. Bioinformatics 2014.
## 
## This message can be suppressed by:
##   suppressPackageStartupMessages(library(circlize))
## ========================================

library(hopach)

## Loading required package: cluster

## Loading required package: Biobase

## Loading required package: BiocGenerics

## 
## Attaching package: 'BiocGenerics'

## The following objects are masked from 'package:dplyr':
## 
##     combine, intersect, setdiff, union

## The following objects are masked from 'package:stats':
## 
##     IQR, mad, sd, var, xtabs

## The following objects are masked from 'package:base':
## 
##     anyDuplicated, append, as.data.frame, basename, cbind, colnames,
##     dirname, do.call, duplicated, eval, evalq, Filter, Find, get, grep,
##     grepl, intersect, is.unsorted, lapply, Map, mapply, match, mget,
##     order, paste, pmax, pmax.int, pmin, pmin.int, Position, rank,
##     rbind, Reduce, rownames, sapply, setdiff, sort, table, tapply,
##     union, unique, unsplit, which.max, which.min

## Welcome to Bioconductor
## 
##     Vignettes contain introductory material; view with
##     'browseVignettes()'. To cite Bioconductor, see
##     'citation("Biobase")', and for packages 'citation("pkgname")'.

#only use the original items and not the average of mindtypes
#turn into z-scores
alldathm <- scale(cleandat[,c(1:6)])

# Color of heatmap
best_col <- colorRamp2(c(-2, 0, 2), c("#82A3FF", "grey", "#8B2E2E"))

#Heatmap with uncentered correlation & clustering with average linkage

uncenter.dist <- function(m){
  as.dist(as.matrix(distancematrix(m, d="cosangle")))
}

row.clus <- hclust(uncenter.dist(alldathm), method = "ave")
col.clus <- hclust(uncenter.dist(t(alldathm)), method = "ave")


Heatmap(alldathm, name = "Mindtype", 
        column_title = "Mindtype Items", column_title_side = "bottom",
        cluster_rows = row.clus, cluster_columns = col.clus,
        col = best_col)

#some splittinggg
Heatmap(alldathm, name = "Mindtype", 
        column_title = "Mindtype Items", column_title_side = "bottom",
        cluster_rows = row.clus, cluster_columns = col.clus,
        column_split = 2, 
        col = best_col)

Heatmap(alldathm, name = "Mindtype", 
        column_title = "Mindtype Items", column_title_side = "bottom",
        cluster_rows = row.clus, cluster_columns = col.clus,
        column_split = 2, row_split = 6,
        col = best_col)

### Heatmap Divided by Survey

#dat5 has different number of rows, cannot add to the other
#random sample of dat5 
set.seed(1234)

dat5random <- dat5[sample(1:nrow(dat5), 300),]


dat2hm <- scale(dat2[,1:6])
dat4hm <- scale(dat4[,1:6])
dat5hm <- scale(dat5random[,1:6])

#Survey 2
row.clus2 <- hclust(uncenter.dist(dat2hm), method = "ave")
col.clus2 <- hclust(uncenter.dist(t(dat2hm)), method = "ave")


hmdat2 = Heatmap(dat2hm, name = "Survey 2", 
        cluster_rows = row.clus2, cluster_columns = col.clus2,
        column_title = "Survey 2",
        col = best_col)

hmdat2

# Survey 4
row.clus4 <- hclust(uncenter.dist(dat4hm), method = "ave")
col.clus4 <- hclust(uncenter.dist(t(dat4hm)), method = "ave")


hmdat4 = Heatmap(dat4hm, name = "Survey 4", 
        cluster_rows = row.clus4, cluster_columns = col.clus4,
        column_title = "Survey 4",
        col = best_col)
hmdat4

# Survey 5
row.clus5 <- hclust(uncenter.dist(dat5hm), method = "ave")
col.clus5 <- hclust(uncenter.dist(t(dat5hm)), method = "ave")


hmdat5 = Heatmap(dat5hm, name = "Survey 5", 
        cluster_rows = row.clus5, cluster_columns = col.clus5,
        column_title = "Survey 5",
        col = best_col)

hmdat5

hmdat2 + hmdat4 + hmdat5

#this was fun but not helpful
#maybe look at them individually survey-wise

Distribution

#promotion vs. prevention orientation
cleandat %>%
  filter(!is.na(Condition)) %>%
  ggplot(mapping = aes(x = promotion, fill = Condition, color = Condition)) +
  geom_density(alpha = 0.3) +
  labs(x = "Average Promotion Mindtype", y = "Distribution") +
  ggtitle("Density Plot of Promotion Mindtype, by Survey") +
  theme_bw()

#not normal distribution


cleandat %>%
  filter(!is.na(Condition)) %>%
  ggplot(mapping = aes(x = prevention, fill = Condition, color = Condition)) +
  geom_density(alpha = 0.3) +
  labs(x = "Average Prevention Mindtype", y = "Distribution") +
  ggtitle("Density Plot of Prevention Mindtype, by Survey") +
  theme_bw()

#what a weird distribution shape

#hmm interesting that prevention frame distribution is not smooth

#difference scores between the mindtypes
#interactive! 
plotdiff <- cleandat %>%
  filter(!is.na(Survey) & !is.na(Condition)) %>%
  ggplot(mapping = aes(x = difference, fill = Survey, color = Survey)) +
  geom_density(alpha = 0.3) +
  facet_wrap(~Condition) +
  labs(x = "Promotion & Prevention Score Difference", y = "Distribution") +
  ggtitle("Density Plot of Mindtypes, by Survey and Condition") +
  theme_bw()

ggplotly(plotdiff)

Group Differences

#setting up 
# Connected Dot Plot yayyy
col_dotplot <- c("#E69F00", "#0072B2", "#D55E00", "#56B4E9", "#800000")

Part of the following R codes for the connected dot plot are from Bowers et al., 2022. ### Politics

cleandat %>%
  filter(!is.na(Politics)) %>%
  group_by(Politics) %>%
  dplyr::summarise(N = n(),
                   "Enhance everyone's health & wellbeing" = mean(pro1, na.rm = TRUE),
                   "Greater economic prosperity to area" = mean(pro2, na.rm = TRUE),
                   "Better future for next generation" = mean(pro3, na.rm = TRUE),
                   "Promotion Average" = mean(promotion, na.rm = TRUE),
                   "Militant/extreme members make it worse" = mean(pre1, na.rm = TRUE),
                   "Don't want loved ones to die/be harmed" = mean(pre2, na.rm = TRUE),
                   "End number of Americans killed/injured" = mean(pre3, na.rm = TRUE),
                   "Prevention Average" = mean(prevention, na.rm = TRUE)) %>%
  pivot_longer(c("Enhance everyone's health & wellbeing",
                   "Greater economic prosperity to area",
                   "Better future for next generation",
                   "Promotion Average",
                   "Militant/extreme members make it worse",
                   "Don't want loved ones to die/be harmed",
                   "End number of Americans killed/injured",
                   "Prevention Average"), 
               names_to = "Item", values_to = "Mean") %>%
  ggplot(aes(x = factor(Item, levels = rev(unique(Item)), ordered = TRUE),
             y = Mean, size = N, fill = Politics)) +
  scale_fill_manual(values = col_dotplot) +
  geom_line(aes(group = Item), size = .2) +
  geom_point(aes(size = N), shape = 21, alpha = .7) +
  theme(legend.key.height = unit(1, 'cm'), legend.key.width = unit(1, 'cm')) +
  ylim(1,4) +
  labs(x = NULL, y = "Mean (Not At All - Extremely Motivating)",
       title = "Mean by Political Groups") +
  theme(plot.title = element_text(size = 5)) +
  coord_flip() + 
  theme_bw()

Race

cleandat %>%
  filter(!is.na(RaceGroup)) %>%
  group_by(RaceGroup) %>%
  dplyr::summarise(N = n(),
                   "Enhance everyone's health & wellbeing" = mean(pro1, na.rm = TRUE),
                   "Greater economic prosperity to area" = mean(pro2, na.rm = TRUE),
                   "Better future for next generation" = mean(pro3, na.rm = TRUE),
                   "Promotion Average" = mean(promotion, na.rm = TRUE),
                   "Militant/extreme members make it worse" = mean(pre1, na.rm = TRUE),
                   "Don't want loved ones to die/be harmed" = mean(pre2, na.rm = TRUE),
                   "End number of Americans killed/injured" = mean(pre3, na.rm = TRUE),
                   "Prevention Average" = mean(prevention, na.rm = TRUE)) %>%
  pivot_longer(c("Enhance everyone's health & wellbeing",
                   "Greater economic prosperity to area",
                   "Better future for next generation",
                   "Promotion Average",
                   "Militant/extreme members make it worse",
                   "Don't want loved ones to die/be harmed",
                   "End number of Americans killed/injured",
                   "Prevention Average"), 
               names_to = "Item", values_to = "Mean") %>%
  ggplot(aes(x = factor(Item, levels = rev(unique(Item)), ordered = TRUE),
             y = Mean, size = N, fill = RaceGroup)) +
  scale_fill_manual(values = col_dotplot) +
  geom_line(aes(group = Item), size = .2) +
  geom_point(aes(size = N), shape = 21, alpha = .7) +
  theme(legend.key.height = unit(1, 'cm'), legend.key.width = unit(1, 'cm')) +
  ylim(1,4) +
  labs(x = NULL, y = "Mean (Not At All - Extremely Motivating)",
       title = "Mean by Racial Groups") +
  theme(plot.title = element_text(size = 5)) +
  coord_flip() + 
  theme_bw()

# wow this is important for the scale
#For prevention item #1 is significantly different for Black participants re: rating
#we need to revise this item

Gender

cleandat %>%
  filter(Gender == c("Female", "Male")) %>%
  group_by(Gender) %>%
  dplyr::summarise(N = n(),
                   "Enhance everyone's health & wellbeing" = mean(pro1, na.rm = TRUE),
                   "Greater economic prosperity to area" = mean(pro2, na.rm = TRUE),
                   "Better future for next generation" = mean(pro3, na.rm = TRUE),
                   "Promotion Average" = mean(promotion, na.rm = TRUE),
                   "Militant/extreme members make it worse" = mean(pre1, na.rm = TRUE),
                   "Don't want loved ones to die/be harmed" = mean(pre2, na.rm = TRUE),
                   "End number of Americans killed/injured" = mean(pre3, na.rm = TRUE),
                   "Prevention Average" = mean(prevention, na.rm = TRUE)) %>%
  pivot_longer(c("Enhance everyone's health & wellbeing",
                   "Greater economic prosperity to area",
                   "Better future for next generation",
                   "Promotion Average",
                   "Militant/extreme members make it worse",
                   "Don't want loved ones to die/be harmed",
                   "End number of Americans killed/injured",
                   "Prevention Average"), 
               names_to = "Item", values_to = "Mean") %>%
  ggplot(aes(x = factor(Item, levels = rev(unique(Item)), ordered = TRUE),
             y = Mean, size = N, fill = Gender)) +
  scale_fill_manual(values = col_dotplot) +
  geom_line(aes(group = Item), size = .2) +
  geom_point(aes(size = N), shape = 21, alpha = .7) +
  theme(legend.key.height = unit(1, 'cm'), legend.key.width = unit(1, 'cm')) +
  ylim(1,4) +
  labs(x = NULL, y = "Mean (Not At All - Extremely Motivating)",
       title = "Mean by Gender") +
  theme(plot.title = element_text(size = 5)) +
  coord_flip() + 
  theme_bw()

Socio-economic Status

cleandat %>%
  filter(!is.na(SES)) %>%
  group_by(SES) %>%
  dplyr::summarise(N = n(),
                   "Enhance everyone's health & wellbeing" = mean(pro1, na.rm = TRUE),
                   "Greater economic prosperity to area" = mean(pro2, na.rm = TRUE),
                   "Better future for next generation" = mean(pro3, na.rm = TRUE),
                   "Promotion Average" = mean(promotion, na.rm = TRUE),
                   "Militant/extreme members make it worse" = mean(pre1, na.rm = TRUE),
                   "Don't want loved ones to die/be harmed" = mean(pre2, na.rm = TRUE),
                   "End number of Americans killed/injured" = mean(pre3, na.rm = TRUE),
                   "Prevention Average" = mean(prevention, na.rm = TRUE)) %>%
  pivot_longer(c("Enhance everyone's health & wellbeing",
                   "Greater economic prosperity to area",
                   "Better future for next generation",
                   "Promotion Average",
                   "Militant/extreme members make it worse",
                   "Don't want loved ones to die/be harmed",
                   "End number of Americans killed/injured",
                   "Prevention Average"), 
               names_to = "Item", values_to = "Mean") %>%
  ggplot(aes(x = factor(Item, levels = rev(unique(Item)), ordered = TRUE),
             y = Mean, size = N, fill = SES)) +
  scale_fill_manual(values = col_dotplot) +
  geom_line(aes(group = Item), size = .2) +
  geom_point(aes(size = N), shape = 21, alpha = .7) +
  theme(legend.key.height = unit(1, 'cm'), legend.key.width = unit(1, 'cm')) +
  ylim(1,4) +
  labs(x = NULL, y = "Mean (Not At All - Extremely Motivating)",
       title = "Mean by Socio-economic Status") +
  theme(plot.title = element_text(size = 5)) +
  coord_flip() + 
  theme_bw()

#even though we have a smaller pool for $165k+
#not surprising that the richest group has the 
#least interest in greater economic prosperity to the area
#also number of people killed (decreased empathy?)

Interaction between Groups

Political Group and Race

#Promotion 
cleandat %>%
  filter(!is.na(Politics) & !is.na(RaceGroup)) %>%
  mutate(difference = promotion - prevention) %>%
  ggplot(mapping = aes(x = difference)) +
  geom_density(fill = "#82A3FF", alpha = 0.5) +
  facet_grid(Politics ~ RaceGroup) +
  labs(x = "Difference between Promotion & Prevention") +
  theme_bw()

### Political Group and Gender

#DV

# Gender and Political Groups

cleandat %>%
  filter(Gender == c("Male", "Female") & !is.na(Politics)) %>%
  group_by(Gender, Politics) %>%
  summarize(DV1 = mean(DV)) %>%
  ggplot(mapping = aes(x = Politics, y = DV1, fill = Politics)) +
  geom_point() +
  ylim(1,5) +
  labs(x = NULL, y = "Willingness to Engage", fill = "Political Group") +
  guides(fill = FALSE) +
  coord_flip() +
  facet_wrap(~Gender) +
  theme_bw()

## `summarise()` has grouped output by 'Gender'. You can override using the
## `.groups` argument.

Intersectionality

First introduced in 1989 by Dr. Kimberlé Williams Crenshaw and adapted by Dr. Patricia Hill Collins a year later, intersectionality refers to how people experience different parts of their identities simultaneously instead of separately. For example, a Black queer woman experiences her life as Black, queer, and as a woman - we cannot fully separate these identities from one another.

cleandat %>%
  filter(Gender == c("Male", "Female") & !is.na(RaceGroup)) %>%
  mutate(difference = promotion - prevention) %>%
  ggplot(mapping = aes(x = difference)) +
  geom_density(fill = "#82A3FF", alpha = 0.5) +
  facet_grid(Gender ~ RaceGroup) +
  labs(x = "Difference between Promotion & Prevention") +
  theme_bw()

#higher scores mean leaning more towards prevention
#don't have enough respondents for other groupings